Open wconnell opened 3 months ago
Hi @wconnell
The code for Figure 5 is located here.
There were some changes made when switching to the final Nature Methods version of the paper, most notably the introduction of a sample correction. Here's an access link for the final version of the paper/figure.
Let me know if this material isn't enough to help and I'll take a closer look at what you posted.
I was unable to recover any perturbations with significant effects using the e-distance and corresponding test in the Norman et al. dataset following paper methods and suggested baselines. Following from the preprint manuscript:
And additional details from methods section:
With min_cells=300 (cell per pert), subsampling/filtering the perturbation conditions resulted in 153 conditions (including control) with 300 total cells. The control cells condition is also subsampled. Upon usage of the e-distance test, 0 passed adjusted significance. I also tried this with min_cells=50 (cell per pert), with the same result of 0 passing significance. This is in contrast to Supp Table 4.
Can you please clarify this discrepancy? Pointing to the code would help. Thanks.
Here is the code I used -